Re: [PATCH] net: meth: Add set_rx_mode hook to fix ICMPv6 neighbor discovery

[Stephen Hemminger <shemminger@vyatta.com>]

On Tue, 27 Dec 2011 17:48:32 -0500
Joshua Kinard <kumba@gentoo.org> wrote:

> On 12/27/2011 17:34, Stephen Hemminger wrote:
> 
> > On Tue, 27 Dec 2011 16:29:57 -0500
> > Joshua Kinard <kumba@gentoo.org> wrote:
> > 
> >> MIPS I/O registers are always memory-mapped, and to prevent the compiler
> >> from trying to over-optimize, volatile is used to make sure we always read 
> >> a
> >> value from the hardware and not from some cached value.
> > 
> > Almost every other network driver had memory mapped register.
> > The problem is volatile is that the compiler is stupid and wrong.
> > Using explicit barriers is preferred and ensures correct and fast
> > code.
> 
> 
> I am somewhat new to driver development, so I do not know all the tricks of
> the trade just yet.  Do you have references to doing explicit barriers that
> I can look at?  Might be worth trying on the RTC driver I have to get the
> hang of them.
> 

Start by reading volatile considered harmful and memory barriers in kernel
Documentation directory. Paul does a better job of explaining it than
I could ever do :-)